Basic Prompting

LLMlight is a library for lightweight, modular and efficient use of LLM and RAG workflows. Below are quick examples using the main functions of the library.

from LLMlight import LLMlight

# Initialize an LLMlight client (default settings)
client = LLMlight(model='mistralai/mistral-small-3.2')

# Ask a question using a language model
response = client.prompt('What is the capital of France?')
print(response)

Working with Files (PDFs)

Add the content of a PDF to memory:

# Import library
from LLMlight import LLMlight

# Initialize model and memory
client = LLMlight(model='mistralai/mistral-small-3.2')
client.memory_init(file_path='knowledge_base.mp4')

# Add a PDF file to the memory (extracts and chunks text automatically)
client.memory_add(files='https://erdogant.github.io/publications/papers/2020%20-%20Taskesen%20et%20al%20-%20HNet%20Hypergeometric%20Networks.pdf')

# Store memory to disk
client.memory_save(overwrite=True)

# Query on the new knowledge
response = client.prompt('Summarize the document.')
print(response)

Create Summaries

Creating summaries can be done using the summary functionality. In this example, a sliding window with the last 5 chunks is kept in memory and expanded.

# Import library
from LLMlight import LLMlight

# Initialize
client = LLMlight(model='mistralai/mistral-small-3.2', top_chunks=5)

# Add multiple PDF files to the database
url = 'https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf'
pdf_text = client.read_pdf(url)

# Create summary
text_summary = client.summarize(context=pdf_text)

---

---